Breast cancer is a major source of morbidity and mortality, and remains the most common cancer occurring in U.S. women, with over 183,000 new cases for 2008 (Ries et al. 2008). This application addresses two aspects of breast cancer development: germline mutations, as in two tumor suppressor genes BRCA1/2, and somatic mutations characterized by copy number changes and allelic loss. The first aim of this application is on the estimation of the age-dependent penetrance function of candidate genes like BRCA1/2 from population-based case-control studies without external disease incidence data. The estimates obtained from the data and the modeling framework will be used to estimate the mutational carrier probability and predict the breast cancer risk of a woman. A shared frailty model will be used to account for the widely observed substantial risk heterogeneity among families. The model will be extended to a general multivariate frailty model to allow for varying degrees of the correlation for different types of relations and multiple candidate gene-related phenotypes. Despite recent successes in identifying novel candidate loci for breast cancer, about three- quarter of breast cancer cases having no clustering in families. It is believed that cancer develops as a result of an accumulation of genetic aberrations at chromosomal locations that are critical in maintaining normal cell functions. The loss and gain of genetic information from specific chromosomal locations are considered an indication for the involvement of tumor suppressor or oncogenes in the tumorigenesis. Using genomic instability data, the second aim of this application is on identifying novel oncogenic networks involved in breast cancer tumor development. Graphical model-based methods will be developed by using novel sparse regression and multiple testing techniques. As the high throughput genotyping technologies become widely available, researchers are able to conduct large scale genome-wide association and genomic instability studies. This allows us to identify novel candidate loci that may have only moderate effect on breast cancer risk or discover oncogenic networks at such a detail that is not previously feasible. There is a great need to characterize these novel genes and provide inference for these networks. The methods proposed here are efforts towards this direction.